Voy is a lightweight WebAssembly-based vector similarity search engine written in Rust that enables semantic search capabilities directly in browser or edge environments. It is designed to operate on embeddings generated from text or other data, allowing applications to perform meaning-based search rather than simple keyword matching. The engine uses efficient data structures such as k-d trees to index vectors and perform fast nearest-neighbor queries, making it suitable for real-time search experiences. One of its key advantages is its extremely small footprint, allowing it to be deployed easily on CDNs or client-side applications without significant overhead. Voy also supports serialization and deserialization of indexes, enabling developers to persist and transfer search data across environments. The architecture is optimized for modern web APIs, including asynchronous execution and compatibility with Web Workers.
Features
- Vector similarity search using embeddings
- k-d tree indexing for fast nearest-neighbor queries
- WebAssembly execution for browser and edge deployment
- Serializable and resumable search indexes
- Small bundle size optimized for CDN delivery
- Asynchronous execution with Web Worker compatibility